Conditional Mutual Information Maximization and Restricted Boltzmann Machine for Stock Price Prediction

Using various machine learning (ML) methodologies in forecasting the stock prices has lately been successful. Nonetheless, most of these models depend on the narrow set of attributes as input and lacks the sufficient information to offer the estimates of stock market. To enhance the stock price prediction models, conditional mutual information maximizing (CMIM) method is used for data preprocessing and feature selection based on restricted boltzmann machine (RBM). By optimizing their conditional mutual information with the target variable, CMIM assists in selecting the most significant characteristics which reduces the dimensionality and duplication. Following that, the picked features are refined using RBM, ML model is capable of detecting the hidden patterns in data. These approaches use selection algorithms to extract the key features. Hence, improves the performance of stock price prediction models. In terms of prediction accuracy, the experiments results indicate the significantly outperforming standard feature selection strategies. This proposed method outperforms with 97.69% accuracy. Thus, the model offers a solid foundation for financial forecasting.